Skill: AI Tool Review

Purpose

Bilal's highest-performing post format. Pick a category, try 5-7 tools, rank with scores, ship a verdict.

When to Use

• Plan slot tagged angle: AI tool review
• A new AI/dev tool launches that fits Bilal's stack (React, .NET, agents, automation)

Process

1. Pick category (e.g., "AI website builders", "code agents", "browser agents"). Avoid categories already covered in last 60 days.
2. List 5-7 tools. Note free-tier availability.
3. For each tool: try the same prompt or task. Score 1-10. Capture one specific observation (speed, output quality, pricing surprise).
4. Pick a winner and a "skip this" loser. Cite both with evidence.
5. Draft post in the comparison format Bilal has used before. Score table inline (text, not image).
6. Add CTA: "İlk yorumda link" if Bilal will drop a link. Otherwise no CTA.

Output Snippets

- Firebase Studio: hızlı ama tasarımda zayıf. 1/10
- Claude: içerik sağlam, görsel zayıf. 3/10
- Lovable: uzun komutlar, canlı düzenleme, tutarlı. 8/10

Verdict: Hız + şık tasarım → Vercel V0. Kararlı geliştirme → Lovable.

Quality Bar

• 5-7 tools minimum, no fewer (single-tool review reads as ad)
• Each tool: a score + a concrete observation
• Winner + loser must each be named
• Personal verdict at the end (not "it depends")

Skill: Analytics Review

Purpose

Score last week's posts against KPIs, name a top and a bottom, push learnings to MEMORY.md.

Inputs

• Bilal pastes weekly LinkedIn analytics export into data/imports/analytics_YYYY-MM-DD.csv (or markdown table)
• Own outputs/ from the same week

Process

1. Match each published post in analytics to its draft. If a draft was never published, flag it.
2. Compute: reactions, comments, reshares, impressions per post. Compute avg + median.
3. Rank posts. Identify top 1 and bottom 1.
4. For top post: extract hook pattern, angle, time. Compare to existing MEMORY.md patterns. If new, add it.
5. For bottom post: extract failure mode. Add as anti-pattern in MEMORY.md if it confirms an existing hypothesis.
6. Write report to outputs/YYYY-MM-DD_analytics.md.

Report Template

# Analytics — Week of YYYY-MM-DD

## Scoreboard
| Post | Reactions | Comments | Impressions | Status |
|------|-----------|----------|-------------|--------|

## Top post
- Slug, hook, why it worked

## Bottom post
- Slug, hook, hypothesized cause

## Trailing KPIs (4-week rolling)
| Metric | Target | 4wk avg |
|--------|--------|---------|

## Memory updates
- <pattern added/removed>

## Next week
- 1-2 specific experiments to run

Quality Bar

• A pattern only enters MEMORY.md after 2 confirming posts
• Anti-pattern goes in after 2 confirming bottom posts
• Report fits on one screen

Skill: Build-in-Public

Purpose

Convert real glossgo engineering events (deploys, incidents, perf wins, architecture decisions) into LinkedIn posts that drive dev/recruiter/investor inbound.

When to Use

• Plan slot tagged angle: build-in-public
• An incident postmortem just landed (agents/incident-commander/outputs/)
• A perf win is recorded (agents/performance-monitor/outputs/)
• A milestone hits (uptime, response time, agent ledger cost, deploy count)

Process

1. Pull the source artifact path. Read it.
2. Extract one number, one cause, one lesson. Discard the rest.
3. Confirm the number is safe to share (no PHI, no customer count, no MRR). Cross-check with RULES.md of kvkk-compliance and investor-relations.
4. Write the post in failure-first or specific-number-first style. Avoid "we crushed it" tone.
5. Optional: invite specific discussion. "What's your p95 on Cloudflare Workers?" type questions outperform "thoughts?".

Templates

Postmortem narrative

Geçen Salı saat 03:00'te <component> patladı.
Sebep: <one line>.
Çözüm: <one line>.
Öğrendim: <one line>.

Şimdi <metric> şu seviyede: <number>.

Perf win

glossgo API p95: <old> → <new> ms.
Tek değişiklik: <one line>.
<Reasoning, 2-3 sentences>.

Quality Bar

• Source artifact path cited in ## Notes
• At most one number in the body — pick the strongest
• No internal codenames, no employee names without consent
• No customer details, even anonymized

Skill: Comment Engine

Purpose

Draft 3-5 comments per day for Bilal to drop on relevant posts in his network. Comments compound reach far faster than posts.

Targeting

• Posts from glossgo team, REM People colleagues, ex-Customertimes/Otokoc/Ford network — priority 1
• Dev/AI builder Turkish audience — priority 2
• Founder/investor accounts Bilal follows — priority 3

Process

1. Bilal provides 5-10 post URLs (manual paste into data/imports/comment_queue.md) OR agent reads a list of monitored authors.
2. For each, draft a 1-3 sentence comment that:
   • Adds a specific data point, tool, or counter-example (no "great post!")
   • Does not pitch glossgo unless the post explicitly invites it
   • Has Bilal's voice (direct, technical, dry humor allowed)
3. Save to outputs/YYYY-MM-DD_comments.md.

Comment Patterns That Compound

• Receipt: cite a number from your own experience that confirms or contradicts the post
• Counter: respectful disagreement with reasoning (highest reach generator)
• Extension: add a tool/library the OP missed
• Question: only if the question opens new territory, not "could you elaborate?"

Banned

• "Great post!"
• "Thanks for sharing"
• "100%" / "Totally agree"
• Anything an LLM would auto-produce

Output Template

# Comments YYYY-MM-DD

## <post URL or author>
**Type**: counter
**Draft**:
> <comment text>
**Why**: <one-line reason this adds value>

Skill: Post Drafting

Purpose

Turn one slot from the weekly plan into a publish-ready draft in Bilal's voice.

Process

1. Read the plan row — pull angle, hook seed, anchor, day/time.
2. Write the hook (line 1, <=80 chars). Test against banned-opener list in RULES.
3. Draft body (<=180 words). One idea. One number if relevant. Numbered or bulleted only if comparing 3+ items.
4. Add CTA only if there is a real ask (try the tool, share your stack, etc.). Otherwise end on a question or a short verdict.
5. Suggest 0-3 hashtags (Turkish if Turkish post). No more than 3.
6. Suggest image brief if the angle benefits visually. Drop a JSON skeleton in the draft for dispatch to ai-visual-studio.
7. Save to outputs/YYYY-MM-DD_post_<slug>.md. Mark status: draft → ready_for_review.

Draft Template

# Post: <slug>
- Status: ready_for_review
- Day/time: <day> <HH:MM TRT>
- Anchor: <number or tool>
- Expected reactions: <N>

## Hook
<line 1>

## Body
<<=180 words>

## Hashtags
#tag1 #tag2

## Image brief (optional)
{"style": "...", "subject": "...", "format": "1:1 LinkedIn"}

## Notes for Bilal
- Why this hook
- Any caveat before publishing

Quality Bar

• Read aloud test: does this sound like a person, not a brand?
• Numbers test: every claim must be backed by a journal entry path or a public source link in ## Notes.
• Banned phrase scan: none of the AI-tell openers from RULES.
• Length test: hook <=80 chars, body <=180 words.

Skill: Post Planning

Purpose

Produce a 7-day post plan with 2-3 slots filled, anchored to upcoming glossgo events, recent agent outputs, and verified hook patterns.

Process

1. Pull last 7 days of journal/ entries — list deploys, incidents, milestones, perf wins.
2. Pull agents/data-bi/outputs/ latest KPI snapshot. Extract 1-2 numbers safe to share publicly.
3. Check knowledge/STRATEGY.md for quarter themes and launch beats.
4. For each post slot (target 2-3 / week):
   • Pick angle: (a) AI tool review, (b) build-in-public, (c) technical opinion, (d) failure / lesson.
   • Pick hook pattern from MEMORY.md that fits the angle.
   • Pick day + time (Tue/Thu/Sat morning default).
   • Define one number or one tool name to anchor the post.
5. Write plan to outputs/YYYY-MM-DD_weekly_plan.md. Leave 1 slot open for reactive content.

Output Template

# Week of YYYY-MM-DD

| Day | Time (TRT) | Angle | Hook seed | Anchor (number/tool) | Image needed? | Status |
|-----|------------|-------|-----------|----------------------|---------------|--------|
| Tue | 09:30 | AI tool review | "Bu hafta X aracını denedim..." | tool name | No | drafted |
| Thu | 10:00 | build-in-public | "glossgo'da bu sabah ..." | p95 number from perf | Yes (architecture) | brief |
| Sat | 09:00 | RESERVED reactive | | | | open |

Quality Bar

• Each filled slot must have an anchor (number, tool, or event). No abstract opinion posts.
• No two consecutive posts in the same angle.
• At least one post per week ties back to glossgo (product, infra, team).

Skill: Profile Rewrite

Purpose

Quarterly full-profile audit + rewrite for Bilal. Aligns headline, About, Experience, Featured, Skills, and custom URL with current glossgo trajectory and the AI/builder positioning.

When to Use

• First-time setup
• Quarterly (every 90 days)
• After a material event (new role, fundraise close, major launch, press hit)

Inputs

• Current LinkedIn HTML/text snapshot in data/imports/profile_snapshot_YYYY-MM-DD.md
• Latest MEMORY.md voice anchors
• knowledge/BRAND.md, knowledge/STRATEGY.md
• Latest analytics report

Process

1. Read current snapshot. Note: headline, About, Featured, Experience, Skills, URL.
2. Score each section against the 2026 algorithm + glossgo positioning. Mark "keep / sharpen / replace".
3. Draft each section per template below.
4. Save to outputs/YYYY-MM-DD_profile_rewrite.md.
5. Bilal updates LinkedIn manually. Log applied changes back to journal.

Section Templates

Headline (220 char cap)

Pattern: <Role @ Company> · <One-line wedge> · <Stack/Specific>. Lead with current role at glossgo, not REM People. Avoid "Exploring" — sounds passive.

Custom URL

Keep linkedin.com/in/bilalsengul. No change.

About (2600 char cap, 7-step)

1. Hook (line 1, short, builder-first)
2. What I'm building right now (glossgo + role + stack)
3. What I'm specifically good at (3-5 bullets)
4. Receipts (numbers from prior roles + glossgo)
5. The opinion / point of view (why I think most beauty-tech / agent-stack work is wrong)
6. Who I want to hear from (devs, builders, recruiters, investors — be specific)
7. CTA line

Featured (4 slots)

• glossgo.com link
• Best technical post (tool review or build-in-public)
• A demo or repo
• A talk / press hit if any

Experience block (glossgo CPO)

• 3-5 lines on what you own
• 3-5 metrics that are safe to publish
• Stack list

Skills (top 5, pinned)

Pick to match the audience you want to attract: investor + builder + recruiter — not the union of every tech you have touched.

Quality Bar

• Headline lead with glossgo role, not REM People
• About passes banned-opener scan
• Every claim verifiable
• Profile reads consistently with last 5 organic posts